PI name/affiliations:

Ann Sheehy, MD, MS

Associate Professor and Division Head, Hospital Medicine

University of Wisconsin Department of Medicine

Potential Co-investigators: Would seek investigators with interest in observation hospital policy to study a representative sample of United States hospitals. This would ideally include urban/rural/community sites, academic/community hospitals, and geographically diverse institutions.

Program overview/introduction: Outpatient (observation) hospital care has grown over the last decade. From 2006-2015, there was a 47.4% increase in outpatient services and 19.5% reduction in inpatient discharges for Medicare beneficiaries according to Medicare Payment Advisory (MedPAC).¹ Medicare’s recent (2013) change to the observation definition, the so-called “2-midnight rule” has resulted in more outpatient encounters (8.1%) in the first year (2014) under the new rule. Compared to 2013, in 2014 there were also more stays of 3 midnights or longer (6%) that did not meet Medicare’s three consecutive inpatient midnight requirement for skilled nursing facility care.²

From a patient perspective, these findings are important because Medicare inpatient hospital stays are covered by Medicare Part A hospital insurance and outpatient observation stays are covered differently under Medicare Part B, which means a patient may pay out of pocket different amounts for the same services depending on whether the hospital stay is considered inpatient or outpatient. Importantly, outpatient nights spent in a hospital do not count towards the 3 consecutive inpatient requirement for skilled nursing facility coverage. From a hospital perspective, Medicare hospital outpatient care has an approximately 10% more negative margin than inpatient.³

Key clinical questions or evidence gaps: Most, if not all, hospitalists encounter patients hospitalized under observation, yet observation remains poorly understood and research is needed.  Barriers to observation research include identifying such claims in administrative data sets, understanding observation in Medicare claims data, dealing with frequent policy change that may impact encounter frequency and type, and possible utilization variability across hospitals or health care systems. The following are key questions/evidence gaps:

1). Are source data valid to accurately identify observation encounters in administrative data sets (i.e. the “OS” Encounter type (ENC_TYPE) variable)?

2). What are differences between Medicare observation and commercial payor observation stays?

3). What types of patients do clinically or financially well under observation, and who is disadvantaged?

4). By patient diagnosis (i.e. lung cancer, renal failure, etc.), what are the outcomes of observation stays?

5). When comparing the same patient type, are there differences between ward based and unit based observation care (i.e. patient selection, social and demographic factors, source of entry into observation encounter, etc.)?

6). What is the source of entry for observation stays? Is it the ED, clinics, outside hospitals? Are there differences in these types of observation encounters?

Aims and Hypotheses:

Primary Aim: To validate observation encounters in administrative data sets, including the “OS” Encounter type (ENC_TYPE) variable in the PCORnet data set, and UHC data.

Rationale: The OS Encounter Type is included in v.3.1 and is new and may be variably reported by sites. In addition, it is not clear how UHC sites report their observation encounters. Based on our experience, identification of observation claims can be challenging based on how encounters are billed and if there was a status change. Understanding what sources are able to report observation encounters accurately and validating the observation variable will create infinite possibilities for further observation research. Additional aims will be outlined based on reliability of the observation encounter variable in these data sets. 

Hypotheses

1). There will be variability in source reporting of observation encounters

2). Hospitals reporting observation encounters will use a similar methodology if using a similar electronic health record (EHR)

Study design: The study design would depend on data sources used, and ability to access primary data and discuss basic information from submitting sites. For example, if using PCORnet data, we would want to first determine what sites submitted encounter type “OS” (observation) and calculate a percent of sites reporting. Depending on number of sites reporting OS encounters, we would then want to ask these sites (or a sample of these sites) what EHR they have and what methodology they used for determining observation stays. A sample chart review of OS encounter type stays, and ideally, a sample review of encounters where status changes from inpatient to observation and vice versa would be ideal.

Characteristics of sites who might participate: This would require more discussion based on sites reporting. As per above, ideally we would have a representative sample of United States hospitals to study which would ideally include urban/rural/community sites, academic/community hospitals, and geographically diverse institutions.

Potential funders: Initially we would fund this with Department of Medicine/Division of Hospital Medicine research development funds and personnel. Assuming this variable can be validated and used, funding for future projects would depend on the project-specific goal. For example, looking at cancer patient outcome after observation hospitalization may prompt submission to an NCI RFA.

1). MedPAC March 2017 Report to Congress: Medicare Payment Policy. Available at:  http://medpac.gov/docs/default-source/reports/mar17_entirereport224610adfa9c665e80adff00009edf9c.pdf?sfvrsn=0. Accessed July 7, 2017.

2). OEI-02-15-00020 Office of Inspector General: Vulnerabilities Remain Under Medicare’s 2-Midnight Hospital Policy. Available at: https://oig.hhs.gov/oei/reports/oei-02-15-00020.pdf. Accessed July 7, 2017.

3). MedPAC March 2015 Report to Congress. Medicare Payment Policy. Figure 3-5. Available at: http://medpac.gov/docs/default-source/reports/mar2015_entirereport_revised.pdf?sfvrsn=0. Accessed July 7, 2017.